Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Preventing Non-intrusive Load Monitoring Privacy Invasion: A Precise Adversarial Attack Scheme for Networked Smart Meters

Jialing He, Jiacheng Wang, Ning Wang, Shangwei Guo, Liehuang Zhu, Dusit Niyato, Tao Xiang

TL;DR

This work addresses privacy leakage in non-intrusive load monitoring by proposing ADV-NILM, a Jacobian-based adversarial attack tailored for time-series NILM. The authors formulate both a standard adversarial NILM problem and a practical variant with a zero-sum perturbation constraint to preserve billing accuracy, and they derive a Jacobian-driven optimization algorithm (jacoc-Adam) to generate imperceptible perturbations. By selecting Seq2seq as the target model and validating on REDD and UK-DALE, they demonstrate large degradation of appliance-level predictions while maintaining correct billing, and show perturbation transfer across multiple NILM models. The approach offers a practical privacy protection mechanism for smart meters with demonstrated transferability and robustness to both white-box and black-box settings, potentially influencing policy and deployment of privacy-preserving schemes in smart grids.

Abstract

Smart grid, through networked smart meters employing the non-intrusive load monitoring (NILM) technique, can considerably discern the usage patterns of residential appliances. However, this technique also incurs privacy leakage. To address this issue, we propose an innovative scheme based on adversarial attack in this paper. The scheme effectively prevents NILM models from violating appliance-level privacy, while also ensuring accurate billing calculation for users. To achieve this objective, we overcome two primary challenges. First, as NILM models fall under the category of time-series regression models, direct application of traditional adversarial attacks designed for classification tasks is not feasible. To tackle this issue, we formulate a novel adversarial attack problem tailored specifically for NILM and providing a theoretical foundation for utilizing the Jacobian of the NILM model to generate imperceptible perturbations. Leveraging the Jacobian, our scheme can produce perturbations, which effectively misleads the signal prediction of NILM models to safeguard users' appliance-level privacy. The second challenge pertains to fundamental utility requirements, where existing adversarial attack schemes struggle to achieve accurate billing calculation for users. To handle this problem, we introduce an additional constraint, mandating that the sum of added perturbations within a billing period must be precisely zero. Experimental validation on real-world power datasets REDD and UK-DALE demonstrates the efficacy of our proposed solutions, which can significantly amplify the discrepancy between the output of the targeted NILM model and the actual power signal of appliances, and enable accurate billing at the same time. Additionally, our solutions exhibit transferability, making the generated perturbation signal from one target model applicable to other diverse NILM models.

Preventing Non-intrusive Load Monitoring Privacy Invasion: A Precise Adversarial Attack Scheme for Networked Smart Meters

TL;DR

This work addresses privacy leakage in non-intrusive load monitoring by proposing ADV-NILM, a Jacobian-based adversarial attack tailored for time-series NILM. The authors formulate both a standard adversarial NILM problem and a practical variant with a zero-sum perturbation constraint to preserve billing accuracy, and they derive a Jacobian-driven optimization algorithm (jacoc-Adam) to generate imperceptible perturbations. By selecting Seq2seq as the target model and validating on REDD and UK-DALE, they demonstrate large degradation of appliance-level predictions while maintaining correct billing, and show perturbation transfer across multiple NILM models. The approach offers a practical privacy protection mechanism for smart meters with demonstrated transferability and robustness to both white-box and black-box settings, potentially influencing policy and deployment of privacy-preserving schemes in smart grids.

Abstract

Smart grid, through networked smart meters employing the non-intrusive load monitoring (NILM) technique, can considerably discern the usage patterns of residential appliances. However, this technique also incurs privacy leakage. To address this issue, we propose an innovative scheme based on adversarial attack in this paper. The scheme effectively prevents NILM models from violating appliance-level privacy, while also ensuring accurate billing calculation for users. To achieve this objective, we overcome two primary challenges. First, as NILM models fall under the category of time-series regression models, direct application of traditional adversarial attacks designed for classification tasks is not feasible. To tackle this issue, we formulate a novel adversarial attack problem tailored specifically for NILM and providing a theoretical foundation for utilizing the Jacobian of the NILM model to generate imperceptible perturbations. Leveraging the Jacobian, our scheme can produce perturbations, which effectively misleads the signal prediction of NILM models to safeguard users' appliance-level privacy. The second challenge pertains to fundamental utility requirements, where existing adversarial attack schemes struggle to achieve accurate billing calculation for users. To handle this problem, we introduce an additional constraint, mandating that the sum of added perturbations within a billing period must be precisely zero. Experimental validation on real-world power datasets REDD and UK-DALE demonstrates the efficacy of our proposed solutions, which can significantly amplify the discrepancy between the output of the targeted NILM model and the actual power signal of appliances, and enable accurate billing at the same time. Additionally, our solutions exhibit transferability, making the generated perturbation signal from one target model applicable to other diverse NILM models.

Paper Structure

This paper contains 28 sections, 16 equations, 3 figures, 7 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. DNN-based NILM models
  4. Adversarial attacks
  5. Preliminaries
  6. NILM Problem
  7. Solution Model for NILM Problem
  8. Threat model
  9. Problem formulation
  10. Adversarial NILM Problem
  11. Practical Adversarial NILM Problem
  12. Adversarial-NILM Solution Model
  13. Overview
  14. NILM Model Description
  15. Perturbed Signal Generation
  16. ...and 13 more sections

Figures (3)

  • Figure 1: Framework of our ADV-NILM. ADV-NILM generates perturbation signals using a trained NILM model, which are subsequently added to the original smart meter readings. The perturbed readings effectively deceive adversaries, causing mainstream NILM models to infer appliance-level power usage that significantly deviates from the actual values.
  • Figure 2: The comparisons between the truth power signal and predicted output from the target NILM model for the fridge.
  • Figure 3: The comparisons between the truth power signal and predicted output from the target NILM model for the dishwasher. (The left column corresponds to the original predicted output and the right column is for the output after the perturbing operation by our ADV-NILM model with practicality consideration.)